Analysis of Coverage and Performance of the Variable Sized Replications Simulation Method in Parallel

Parallel simulation methods can be used to reduce the execution time of the simulations of complex systems. In this context, we apply the method of the replications in parallel in order to reduce the execution time of the models. The particular technique consists of starting a fixed number of replications that are maintained in execution until a stop criterion is achieved. Different stopping criteria can be used in order to finish the simulation. These criteria must be robust in order to provide results according the applied statistical theory. The results obtained with two habitual criteria applied to a simple simulation model show that it is necessary to provide more elaborated stopping criteria of the simulation in order to obtain the adequate coverage. In order to reduce the detected problems we propose a new stopping criterion that increments the coverage of the true mean. The analysis of performance shows that, obviously, this better coverage has the cost of more simulation time, but it is important to get results statistically correct.